Accelerating the Production of Synthetic Seismograms by a Multicore Processor Cluster with Multiple GPUs

In this work we propose two different parallel versions of the software package COMPSYN, devoted to the production of syntethic seismograms. The first version consists in the parallelization of the code to run on a cluster of multicore processors and is obtained by exploiting the MPI paradigm and OpenMP API to the end of maximizing the performance on multicore processors. The second version exploits the set of GPU associated to the multicore processor cluster and uses CUDA to take advantage of the GPU's computational power. We analyze the application performance of the two different implementations by using a real case study. In particular, we obtain for the GPU version a speedup of 10x over the parallelized version running on the cluster of multicore processors. Furthermore, we can estimate about at least 100x the speedup of the GPU version using a single node of the cluster with respect to the original sequential version.

Collaboration
Agreement between Department of Computer Science, Sapienza University
of Rome and the Istituto Nazionale di Geofisica e Vulcanologia,
Rome, Italy, 2011. project n.
C26G074ABJ.

[1] MPI: A Message-Passing Interface standard. http://
www.mpi.org.
[2] The OpenMP API specification for parallel programming.
http://openmp.org.
[3] CPTI Working Group, Catalogo parametrico dei terremoti
italiani, version 2004 (CPTI04). http://emidius.mi.
ingv.it/CPTI04/, 2004. INGV, Bologna.
[4] NVIDIA CUDA C Best practices guide, version 3.2, 2010.
[5] NVIDIA CUDA C Programming guide, version 3.2, 2010.
[6] R. Abdelkhalek. ´ Evaluation des acc´el´erateurs de calcul
GPGPU pour la mod´elisation sismique, 2007. Master thesis,
ENSEIRB, Bordeaux, France.
[7] R. Abdelkhalek, H. Calandra, O. Coulaud, J. Roman, and
G. Latu. Fast seismic modeling and Reverse Time Migration
on a GPU cluster. In High Performance Computing &
Simulation. IEEE, 2009.
[8] M. Geveler, D. Ribbrock, D. G¨oddeke, P. Zajac, and
S. Turek. Efficient finite element geometric multigrid
solvers for unstructured grids on GPUs. In PARENG, 2011.
[9] D. G¨oddeke, R. Strzodka, J. Mohd-Yusof, P. McCormick,
S. H. M. Buijssen, M. Grajewski, and S. Turek. Exploring
weak scalability for FEM calculations on a GPU–enhanced
cluster. Parallel Computing, 33(10–11):685–699, 2007.
[10] D. Kirk and W. Hwu. Programming massively parallel processors.
Morgan Kaufmann Publishers, 2010.
[11] D. Komatitsch, D. G¨oddeke, G. Erlebacher, and D. Mich´ea.
Modeling the propagation of elastic waves using spectral elements
on a cluster of 192 GPUs. Computer Science - Research
and Development, 25(1–2):75–82, 2010.
[12] D. Komatitsch, D. Mich´ea, and G. Erlebacher. Porting a
high-order finite-element earthquake modeling application
to NVIDIA graphics cards using CUDA. J. Parallel Distrib.
Comput., 69:451–460, 2009.
[13] D. Mich´ea and D. Komatitsch. Accelerating a threedimensional
finite-difference wave propagation code using
gpu graphics cards. Geophys. J. Int., 182:380–402, 2010.
[14] P. Micikevicius. 3D finite difference computation on GPUs
using CUDA. In Workshop on General Purpose Processing
on Graphics Processing Units. ACM, 2009.
[15] C. Nugteren. Improving CUDAs Compiler through the Visualization
of Decoded GPU Binaries, 2009. Master thesis,
Eindhoven University of Technology.
[16] A. Olson, J. Orcutt, and G. Frazier. The discrete wavenumber/
finite element method for synthetic seismograms. Geophys.
J. Int., 77:421–460, 1984.
[17] L. Smith. Mixed Mode MPI/OpenMP Programming. Edinburgh
Parallel Computing Centre, 2000.
[18] P. Spudich and R. Archuleta. Techniques for earthquake
ground-motion calculation with applications to source parameterization
of finite faults, pages 205–265. Seismic
Strong Motion Synthetics, B. A. Bolt, 1987.
[19] P. Spudich and L. Xu. Documentation of software package
COMPSYN sxv3.11: programs for earthquake ground motion
calculation using complete 1-D Greens functions. International
Handbook of Earthquake and Engineering Seismology,
2002.